Effects of metformin on head and neck cancer: an umbrella review

Abstract

Background

Metformin is a commonly used antihyperglycemic drug to treat type 2 diabetes mellitus worldwide. However, its effectiveness was recently documented beyond diabetic treatment, such as in the case of cancer. Recent epidemiological evidence suggests that the use of metformin has shown benefits in breast and colorectal cancer; however, its effectiveness was inconclusive in the case of head and neck cancer. This umbrella review aimed to evaluate the effectiveness of metformin among head and neck cancer patients.

Methods

Three electronic databases and additional sources (Google Scholar and citation screening) were used to extract relevant research records. A multistage screening was performed by two independent reviewers, and the study selection process was reported according to PRISMA guidelines. Along with descriptive findings and overlap assessment, a meta-analysis was performed using a random-effects model with 95% confidence interval.

Result

A total of five studies were included. Analysis revealed that metformin increases overall survival probability among head and neck cancer patients by 0.74 times [95% C.I. (0.67–0.81)] compared to non-metformin user, indicating pharmacological effectiveness. Overall heterogeneity was negligible (I2 = 0%). Egger’s test suggested funnel plot asymmetry, indicating potential publication bias. All studies have shown effectiveness in lowering incidence risk.

Conclusion

Findings showed promising potential of metformin among HNC user. However, as there were very few studies available and quality assessment showed low confidence, a more comprehensive evaluation of metformin in diabetic and non-diabetic patients should be conducted to confirm these findings.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-026-16057-4.

Background

Metformin is one of the most common antihyperglycemic drugs used to treat type 2 diabetes mellitus (DMT2) [1]. Recently, a series of epidemiological studies has shown that metformin can reduce the risk of cancer [2–4]. These speculations have sparked a new area of research, where their pharmacodynamic effects have been extensively studied in the treatment of chronic ailments such as cancer. The relation between DMT2 and cancer was mainly illustrated by “hyperinsulinemia and hyperglycemia” [5]. In DMT2 patients with insulin resistance, activation of multiple pro-inflammatory cytokine pathways increases, thereby increasing the risk of mitogenic activity [6]. With the use of metformin, endocrine metabolism of the human body can be altered, which can support the anti-tumorigenic behavior within the cells [7].

With focus shifting beyond the DMT2 horizon, metformin use has been increasingly tested against different forms of cancer, with the most common being breast cancer, colorectal cancer, pancreatic cancer, and, very recently, head and neck cancer (HNC) [5, 8]. Evidence in context of breast cancer, pancreatic cancer and colorectal cancer suggests that metformin lowers the insulin and insulin-like growth factor-1 (IGF-1) level by activating AMPK (adenosine monophosphate-activated protein kinase), which reduces cancer cell proliferations [1, 9, 10]. It was further reported by few studies that it also impact pathways like mTOR (mechanistic target of rapamycin) and STAT-3 (signal transducer and activator of transcription 3) which are important mechanisms for cancer cell growth and survival [6, 9, 11]. In contrast, efficacy studies on metformin and cancer yielded conflicting results. For instance, the MA.32 study in phase III trials (randomized controlled) for breast cancer showed that the intervention group, who were given adjuvant metformin (drug dosage = 850 mg), did not show any improvement in disease-free survival over a period of 5 years [11]. As tumor hypoxia was found to be strongly associated with radiotherapy resistance and the addition of metformin during treatment showed otherwise (improved tumor oxygenation), its efficacy becomes debated [12]. Patients with non-small lung cancer were also investigated with a combination of metformin and chemotherapy; the results, however, did not show any additional benefits of using metformin as an anti-tumorigenic drug [13, 14].

While evidence on the overall usefulness of metformin in cancer patients is limited, evidence in the context of HNC is even more limited. As oral and pharyngeal cancers were the sixth most common types of cancer in the world, oral squamous cell carcinoma (OSCC) represents more than 90% of the oral cancers. Early evidence from retrospective cohort studies suggests that incorporating metformin into standard HNC treatment can improve overall survival and reduce incidence; however, the effectiveness remains inconclusive and debatable. Therefore, this umbrella review was planned to highlight the effectiveness of metformin use among HNC patients, using the highest level of available scientific evidence. Furthermore, this review will also assess the overall methodological quality of the included evidence.

Methods

Protocol registration

A priori protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO; registration ID: CRD420261299771) [15]. Findings of this review study were reported in accordance with the guidelines of the Joanna Briggs Institute (JBI).

Focused review research question

A research question was developed based on the PICO framework: “Among HNC-diagnosed patients (population), is the use of metformin (intervention) effective in improving their overall survivability and reducing incidence risk rates (outcome)?”

Search databases

Three electronic databases PubMed, Scopus, and Embase were used to extract the most relevant research records for this review. Additionally, the Google Scholar database was used to extract gray literature (as the results were very large, only the first 50 pages were screened) and to perform citation screening of potentially eligible results to track any relevant research records. The records were searched until 10th January 2026, and no language restrictions were kept including eligible research records. However, if a research record was found in a non-English language, systematic translation into English was carried out according to the methodology described by Balk et al. (2012) [16].

Search strategy

A comprehensive search strategy was built around the PICO framework: Population (P) -HNC-diagnosed patients; Intervention (I) - Administration of metformin; Comparison (C) - As reported in the included studies; Outcome (O) - Improvement in overall survival probability and reduction in incidence risk. Search strings were developed using Medical Subject Headings [MeSH] key terms with additional fields such as “Title/Abstract or All Fields” with filters such as systematic review (SR) and meta-analysis (MA), to narrow down the search strategy. Boolean operators, such as “ALL FIELDS”, “AND”, and “OR”, were extensively used to combine search terms as needed. A detailed search strategy was listed in Table 1 for PubMed. For other databases, search strategies are listed in Supplementary Table S1 of the Supplementary Sheet.

Table 1.

Search Strategy

Database

Search Strategy

PubMed

(((((“head and neck neoplasms“[MeSH* Terms] OR “head and neck neoplasms/analysis“[MeSH Terms]) AND (“metformin“[MeSH Terms] OR “metformin/adverse effects“[MeSH Terms] OR “metformin/pharmacology“[MeSH Terms])) AND (“treatment outcome“[Title/Abstract])) AND (“systematic reviews as topic“[MeSH Terms] OR “systematic reviews as topic/methods“[MeSH Terms]); (“head and neck neoplasms“[MeSH Terms]) AND (“metformin“[MeSH Terms] OR “metformin/administration and dosage“[MeSH Terms] OR “metformin/adverse effects“[MeSH Terms]); (((“Head“[Mesh]) OR “Neck“[Mesh]) AND “Treatment Outcome“[Mesh]) AND “Metformin“[Mesh]; ((((“Metformin“[Mesh] OR “Sitagliptin Phosphate, Metformin Hydrochloride Drug Combination“[Mesh]) AND ( “Drug-Related Side Effects and Adverse Reactions“[Mesh] OR “Physiological Effects of Drugs“[Mesh] OR “Metabolic Side Effects of Drugs and Substances“[Mesh] OR “Long Term Adverse Effects“[Mesh] )) AND “Head“[Mesh]) AND “Neck“[Mesh]) AND ( “Systematic Review” [Publication Type] OR “Systematic Reviews as Topic“[Mesh] ); (((((“metformin“[MeSH Terms] OR “metformin/adverse effects“[MeSH Terms] OR “metformin/analysis“[MeSH Terms])) AND (“head and neck“[Title/Abstract])).

* MeSH: Medical Subject Headings

Eligibility criteria

The retrieved research records were only considered for inclusion if:

1. Studies should exclusively focus on HNC.

2. The study has been a systematic review and/or meta-analysis (SR/SRMA) in its design.

3. Studies focusing on metformin as an intervention given to the HNC-diagnosed patients.

4. Studies were published in peer-reviewed journals.

However, all those research studies that did not focus on metformin as an intervention (exclusively) were excluded. Furthermore, study designs such as scoping reviews, rapid reviews, and criteria reviews were excluded from this study.

Study selection

The search results were imported into Rayyan AI for screening [17]. Two independent reviewers (V.M. and J.F.) screened the research records using a multiple-stage screening process. All the potentially eligible studies were further evaluated against the study’s eligibility criteria. Shortlisted SR/SRMAs were further screened based on the full text to be included in the final review. Cohen’s kappa for the level of agreement between the independent authors was 0.92. Any conflict raised during the screening process was resolved through mutual discussion and consensus with the third reviewer.

Assessment of the overlap of primary studies

SRs and SRMAs included in this review were evaluated for overlap among primary studies by constructing a citation matrix (overlap percentage, covered area, corrected covered area), based on the methodologies described by Pieper et al. (2014) and Tewari et al. (2024) [18, 19]. The matrix was defined as:

1. Overlap Percentage: Proportion of the number of studies shared by at least two SR/SRMAs.

2. Covered Area (%): It was calculated as the total number of publications (considering only research results related to the effectiveness of metformin in HNC patients reported in the included SRs/SRMAs) divided by the product of the total number of studies and the total number of SRs.

3. Corrected Covered Area (%): It was calculated by subtracting the number of unique studies from the total number of publications in each review and dividing the result by the difference between the total study count and the product of the total number of studies and SRs.

This analysis was also performed by two reviewers with good agreement (Cohen’s kappa ranging from 0.86 to 0.94). The result was interpreted according to the classification in Table 2. The overall overlap assessment was interpreted based on the corrected covered area (which systematically eliminates duplication in results), and the other two matrices were used to describe the length and breadth of the overlap among the studies.

Table 2.

Citation matrix interpretation percentage

Interpretation	Corrected Covered Area %
Mild Overlap	0–5
Moderate Overlap	6–10
High Overlap	11–15
Very high overlap	> 16

Data extraction, data management and data analysis

A systematic data extraction form was prepared using an MS Excel spreadsheet (version 2402) in accordance with the Joanna Briggs Institute (JBI) guidelines. The data were extracted on 10th January 2026 by two independent reviewers using a standard data extraction form that included: study ID (year), region, registration, guidelines, databases, study design, final included studies, time frame of included studies, language restriction, study sample, overall sample size, meta-analysis (if yes, heterogeneity), and ROB. Furthermore, intervention characteristics were also reported to stratify the effectiveness of metformin in HNC patients.

In addition, a meta-of-meta-analysis was performed to assess the overall survival probability of HNC-diagnosed patients using metformin. Data for this analysis was extracted from those primary studies only which were identified as unique studies after overlap assessment was performed. This also means that effect estimates were directly sourced from unique primary studies (identified from included studies) and not from the published meta-analyses to avoid any duplication of published result. This prevents over-representation of results favoring a particular outcome and avoid any ‘double counting’ to ensure data independence. The analysis was performed in Review Manager 5.4, using a random-effects model with 95% Confidence Interval (C.I.).

Outcomes measured

The primary outcome of this review was to investigate the odds of overall survival probability (time for which patient remains alive from the point of diagnosis or likelihood of survival post HNC diagnosis) among HNC patients treated with metformin. Furthermore, the secondary outcome explores the methodological robustness of the included studies. Additionally, incidence risk rate was also reported in the result which mainly refers to the occurrence of new cases of HNC among populations at risk during the study period. Furthermore, the study focuses on two types of comparison: diabetic but may or may not using metformin as a blood-sugar controlling drug.

Quality assessment

The included studies were assessed for the overall quality of the presented evidence using the Assessment of Multiple Systematic Reviews 2 (AMSTAR 2) critical appraisal tool. This tool included eight critical and eight noncritical domains, which evaluate the overall confidence of each SR. Responses were recorded as “Yes”, “Partial Yes”, or “No”, indicating the overall quality of the included article as “critically low”, “low”, “moderate”, or “high”. Any disagreement was resolved through mutual discussion and with the assistance of a third reviewer (A.M.). The level of agreement between them was 0.92 for different sections.

Results

Study selection

A total of 279 research records were retrieved across three databases (n = 243), cross-reference screening (n = 7), and Google Scholar (n = 29; used only for the retrieval of grey literature). Before the first screening, 76 duplicates were removed after combining the metadata from three electronic databases. A total of 167 articles were considered for the first screening based on only titles and abstracts. Of 167 articles, 155 were removed, leaving only 12 for full-text screening. After the full-text screening, only five articles were selected for inclusion in the final review [20–24]. A detailed PRISMA flowchart was presented in Fig. 1. A detailed list of excluded articles was mentioned in Table S2 of the supplementary document.

Fig. 1.

PRISMA Flowchart for Study Selection. Flow diagram illustrating identification, screening, eligibility assessment, and inclusion of systematic reviews and meta-analyses evaluating the effects of metformin on head and neck cancer, including database searches, Google Scholar, and citation screening, with reasons for exclusion at each stage

Study characteristics

Study characteristics were presented in Table 3. A total of five studies were included in this review. Two of the five studies were from Brazil, and the remaining were from the USA, China, and Spain (n = 1 each). The analysis revealed that none of the studies were registered with PROSPERO or any other public repository to illustrate the review protocol. Two studies [20, 21] cited that “because they followed PRISMA guidelines, the protocol was not registered with any public repository or PROSPERO”. This indicated a potential for bias in the included studies. However, all studies indicated that they followed PRISMA guidelines for reporting their findings. Furthermore, studies extensively used various databases to identify the most relevant research records on this topic, with PubMed and the Cochrane Library being most searched. A total of 27 primary studies were reported by the included SR/SRMAs, published between 2011 and 2021, and focused on evaluating the effectiveness of metformin in HNC patients. Also, only one study [20] showed no language restriction, whereas others focused mainly on English-language articles except Herrán et al. in 2018 [22] which also focused on Spanish language published articles. This might have resulted in a few non-English research studies being excluded from the SRMAs. Additionally, it was observed that studies focused on diabetic patients diagnosed with HNC to assess metformin’s impact on the cancer, hence suggesting non-diabetic but HNC-diagnosed patients were selectively excluded. Most of the included studies were cohort (both retrospective and prospective), followed by observational studies and RCTs. The cumulative sample size was over 1,627,351 as calculated from the included studies. Furthermore, only three studies [21, 23, 24] reported meta-analyses, with heterogeneity ranging from 32% to 96.6%, suggesting low to moderately high heterogeneity. Risk of bias assessment was also performed for all included studies, except two [20, 24].

Table 3.

Study Characteristics

Study ID (Year)	Region	Registration	Guidelines	Databases	Study Designs Included	Final Studies Included	Time frame of included primary studies	Language Restriction	Study Sample	Overall Sample Size	Meta-Analysis (If Yes, Heterogeneity)	RoB1
Rêgo et al. (2015) [20]	Brazil	Not Registered	PRISMA2	Cochrane Library, EMBASE, MEDLINE, LILACS3, PubMed	Cohort– Retrospective and Prospective Studies	3	2012–2014	Yes (only English)	Patients newly diagnosed diabetes and treated with metformin	33,373	No	Not mentioned clearly
Rêgo et al. (2017) [24]	Brazil	Not Registered	PRISMA	Cochrane Library, Embase, LILACS, MEDLINE, PubMed	In vitro (only studies with human subjects)	11	2011–2015	No	Patients’ cell from HNSCC4	Not mentioned clearly	No	GRADE5
Herrán et al. (2018) [21]	Spain	Not Registered	PRISMA	MEDLINE (PubMed), IBECS, LILACS, Cochrane Central Register for Controlled Trials	RCT, Observational studies	13	2012–2018	Yes (only English or Spanish)	Patients with HNC	370,696	Yes [32–62%]	NOS6
Zhang et al. (2021) [22]	China	Not clearly mentioned	PRISMA	PubMed, Embase, China National Knowledge Infrastructure	Cohort studies	67 [overall], in context of HNC total studies included were 5	2007–2019	Yes (only English)	Patients diagnosed with HNC	1,221,725	Yes [96.6%]	Not performed
Jiao et al. (2022) [23]	China	Not Registered	PRISMA	Cochrane Library, Embase, LILACS, MEDLINE, PubMed	RCT7, Observational studies	11	2014–2020	Yes (only English)	Patients with HNC	1557	Yes [37–71%]	NOS

1: RoB Risk of Bias, 2: PRISMA Preferred Reporting Items for Systematic reviews and Meta-Analyses, 3: LILACS Literatura Latino Americana em Ciências da Saúde, 4: HNSCC Head and neck squamous cell carcinoma, 5: GRADE Grading of Recommendations, Assessment, Development and Evaluation, 6: NOS Newcastle-Ottawa Scale, 7: RCT Randomized controlled trials

Assessment of overlap based on the included primary studies in the SRMAs

Overall, the assessment highlighted moderate overlap (9.25%) among the primary studies included in the SRMAs considered in this review. Furthermore, the citation matrix showed that overlap among the primary studies included by these reviews was 25.93%, whereas the unadjusted covered area was 27.40%. Moreover, among the 27 primary studies (relevant to the topic and included in the five SRMAs), more than 30% (n = 18/27) were unique, suggesting that this study does not favor one outcome over the other. A detailed matrix was mentioned in Table S3 of the supplementary document. Additionally, Fig. 2(a) and (b) illustrate the geographical distribution of all the primary studies included in the SRMAs. It can be observed that the majority of the studies were conducted in the USA, China, and Taiwan compared to only one to three publications from other parts of the world (Fig. 2).

Fig. 2.

Geographical distribution of the primary studies identified in the included SRMA. Panels (a) and (b) showing the geographical spread and frequency of primary studies included in the systematic reviews/meta-analyses on metformin and head and neck cancer, highlighting higher concentrations of studies from the USA, China, and Taiwan compared to other regions

Intervention characteristics

Included studies highlighted the effectiveness of metformin among HNC-diagnosed patients, as summarized in Table 4. The intervention characteristics showed that events and controls were mainly categorized as patients with HNC taking metformin or those with HNC but not taking metformin, suggesting it was the exposure variable. Median age of the patients reported by two of five included studies only [20, 24], which was mainly between 53 and 74 years (approximately). This suggests that the targeted patient group was mostly late middle-aged and elderly individuals. Furthermore, HNC was mainly squamous cell carcinoma of the oral cavity and, in some cases, the larynx. However, the outcomes assessed varied depending on the objectives of the studies. Most included studies have evaluated incidence risk and overall survival probabilities; however, few also evaluated the cellular and metabolic levels of metformin’s effectiveness [cell viability, cell apoptosis, cancer recurrence, cell biophysical changes] among HNC patients. Results from the studies showed that those who take metformin had 2.25 times better outcomes compared to non-users, thereby inhibiting cell proliferation (G0/G1 cell cycle arrest and apoptosis) and regulating proteins involved in carcinogenesis pathways. One study by Herrán et al. [22], also highlighted that the incidence of HCN was 92.7/100,000 person per year in the metformin group and 163.6/100,000 person per year in the non-metformin group (13% higher). Also, the quantitative summary indicated that the odds of adverse outcomes were lower among metformin users than among non-users.

Table 4.

Intervention Characteristics

Study ID (Year)	Population Characteristics			Events and Control (n; exposure)		Result [at 95% C.I1.]			Outcome Assessed	Result
	Classification of Patients	Age (Median in years)	Type of Cancer	Events	Controls	OR2/HR3/RR4	p-value	Any Other Metrics
Rêgo et al. (2015) [20]	Diabetic and non-diabetic patients with HNC5	63	Laryngeal	43 (metformin)	162 (non-metformin)	OR = 2.25 (0.9–5.6)	< 0.001	IRR6= 0.53	Incidence risk ratio and survival probability	Diabetic patients with metformin usage showed 2.23 times better outcomes compared to non-users.
Rêgo et al. [24]	Patients’ cell from HNSCC7	Not mentioned clearly	Different oral regions	metformin users	Non-metformin treatment	N/A8	N/A	N/A	Cell viability, apoptosis, cell cycle arrest and regulation of protein expression levels	Metformin was found to be important in inhibiting cellular and metabolic activities that promote cancer cell growth (cell proliferation, inducing G0/G1 cell cycle arrest and apoptosis)
Herrán et al. (2018) [21]	Patients’ with HNSCC	Not mentioned clearly	laryngeal cancer, others but unclassified with LRR9	metformin users	Non-metformin treatment	RR = 0.71 (0.61–0.84)	0.08	N/A	Risk of cancer in the oral cavity, pharynx and larynx; > 5 years survival; and recurrence of primary lesion, cancer sub-site-specific risk, dose-response relationship related to the use of metformin	Findings indicated incidence of HCN among met + 10 (92.7/100,000 p-y) was lower compared to met-11 (163.6/100,000 p-y)
Zhang et al. (2021) [22]	Patients with multiple cancer sites including HNC	N/A	N/A	Metformin ever users	Metformin never users	OR = 0.55 (0.38–0.79)	< 0.001	N/A	Estimate the effects of metformin on cancer risk with other ADMs12	Metformin may be a independent protective factor for cancer risk in T2DM13 patients
Jiao et al. (2022) [23]	Patients’ with HNSCC	Between 53.50 to 74.55	N/A	Metformin users	Non-metformin treatment	Without HNCC (HR) = 0.87 (0.76–0.99)	0.04	N/A	Implications of metformin usage in HNC patients	Metformin had a significant improvement in overall survival of HNC patients, thus supporting metformin as an adjunct to the treatment of HNC.
						With HNCC = 0.67 (0.40–1.09)	0.11

1: C.I Confidence Interval, 2: OR Odds Ratio, 3: HR Hazard ratio, 4: RR Risk ratio, 5: HNC Head and neck cancer, 6: IRR Incident risk ratio, 7: HNSCC Head and neck squamous cell carcinoma, 8: N/A Not available, 9: LRR Loco regional-recurrence, 10: met+ Metformin user, 11: met- Metformin non-user, 12: ADMs Anti-diabetic medicines, 13: T2DM Type 2 diabetes mellitus

Effectiveness of metformin among HNC patients

Incidence risk of HNC among diabetic patients

Three of the five included studies discussed the incidence risk of HNC among metformin user and non-metformin user. In studies by Rêgo et al. (2015) [20] and Rêgo et al. (2017) [21], the odds of HNC was 0.74 times lower (p < 0.01) among cohorts using metformin than among those not using the drug. The study further reports that this incidence risk was lower among HNC patients aged 40–65 years (0.70, p < 0.01) and those aged > 65 years (0.53, p < 0.01) compared to those not taking metformin in the same age subgroups. Furthermore, these studies also mention that the incidence risk of nasopharyngeal cancer was 0.08% (adjusted hazard ratio [AHR = 0.50, 95% CI (0.31–0.80)] compared to oropharyngeal cancer, which was found to be 0.03% [AHR = 0.36, 95% CI (0.17–0.74)]. This suggests that oropharyngeal cancer has a lower incidence risk compared to nasopharyngeal cancer among metformin users. In fact, 76% of patients taking metformin were alive, compared to only 41% who were DMT2-positive but not taking metformin. In addition, a study by Herrán et al. (2018) [22] reported that patients with long-term metformin use had lower odds of laryngeal cancer (OR = 0.41, 95% CI 0.17–1.03) at later ages. Similarly, sub-sites like pharynx and larynx, the risk of cancer incidence was 54% times lower among DMT2 patients using metformin [OR = 0.54, 95% CI (0.29–0.99)]. Interestingly, this study has shown that the incidence risk of HNC cancer was also low among current smokers [OR = 0.13, 95% CI (0.04–0.44)] and those who were consuming > 40 mg alcohol/day (OR = 0.31, 95% CI 0.11–0.88) when compared to non-DMT2 positive participants. Overall, meta-analysis from Jiao et al. (2022) [24] showed that metformin use was significantly associated with decreased risk of HNC (RR = [0.71, 95% CI (0.61–0.84)] suggesting it reduces the risk of HNC incidence among diabetic patients. However, these findings must be interpreted with caution as smoking, age and sex can confound the overall association between metformin consumption and incidence of HNC among diabetic patients.

Overall survival probability among HNC-diagnosed DMT2 patients with/without use of metformin

Three of the five included studies discussed the survival probability of HNC among DMT2 positive patients with/without metformin use. In a study by Rêgo et al. (2015) [20], it was reported that the odds of overall survival among metformin user were 1.7 times higher (OR = 1.77) compared to non-metformin user suggesting better survivability. A meta-analysis by Herrán et al. (2018) [22] showed that survival from HNC among metformin users (RR = 1.71; 95% CI 1.20–2.42) was better than among non-metformin users. Similarly, a study by Jiao et al. (2022) [24] found that overall survival probability among non-metformin user with HNC was 13% higher compared to metformin user [AHR = 0.87, 95% CI (0.76–0.99), p = 0.04]. However, this study also speculates that, due to significantly smaller cohorts in the included primary studies, the effectiveness on overall survivability needs further evaluation. Hence, our review performed a meta-of-meta-analysis to illustrate the effectiveness of metformin on survival probability in HNC patients. This meta-analysis includes unique primary studies identified from overlap assessment from four SRMAs, with a sample size of 3551 HNC diagnosed patients. Figure 3 highlights the odds of survival between DM (diabetes mellitus) patients consuming metformin (met+) or non-metformin user (met-). This meta-analysis includes unique primary studies identified through overlap assessment across four SRMAs, with a sample size of 3551 HNC-diagnosed patients. Figure 3 highlights the odds of survival between DM (diabetes mellitus) patients using metformin (met+) or non-metformin user (met-). The analysis shows that the likelihood of overall survival among metformin users was 26% higher than among non-metformin users [OR = 0.74, 95% CI (0.67–0.81)], indicating the pharmacological effectiveness of the drug.

Fig. 3.

Forest plot between DM (met+) and DM (met-). Forest plot from the meta-of-meta-analysis comparing overall survival in diabetes mellitus patients with head and neck cancer treated with metformin (met+) versus non-metformin users (met−), presenting pooled odds ratio, 95% confidence intervals, and negligible heterogeneity (I² = 0%)

Furthermore, heterogeneity analysis was negligible (I² = 0%) suggesting homogeneity in the study level characteristics (tau² = 0, Chi² = 2.66, df = 6, p>0.05). Additionally, Egger’s test suggested funnel plot asymmetry [− 0.83, 95% CI (− 1.19 to − 0.47); p = 0.02], which suggests inconsistency with small-study effects, indicating higher potential of publication bias (Fig. 4).

Fig. 4.

Publication Bias. Funnel plot assessing small-study effects for the meta-of-meta-analysis of overall survival among metformin users and non-users, demonstrating asymmetry consistent with potential publication bias as supported by Egger’s test

Effect on Disease-Free Survival (DFS) and Loco Regional Recurrence (LRR) among HNC

Three of the five included studies discussed the effect of metformin on DFS and LRR among HNC patients. Rêgo et al. (2015) [20] mentions that a risk of LRR was significantly lower in metformin user compared to their controls [non-metformin user, p = 0.04 and 0.01, respectively). Furthermore, it was mentioned in this study that the 5-year overall survival rate was 87% in patients treated with metformin, compared to 41% in the remaining patients (p = 0.04). However, in a study by Herrán et al. (2018) [22] the association between metformin use and its effects on DFS and loco regional recurrence was not statically significantly. However, this study showed an important trend among metformin users which illustrated possible increase in DFS in DM patients using metformin compared to the ones who were not using this drug (OR = 1.99 95% CI 0.82–4.83). Furthermore, this study also reports that LRR was observed to be reduced in metformin users and was statistically significant (p = 0.04). The quantitative synthesis performed by this study showed that as non-metformin users had higher HNC related hazard risk of 33% compared to metformin users (HR = 0.67, 95% CI: 0.40–1.09, p = 0.11).

Cellular level of action

Only two of the five included studies discussed the cellular-level effects of metformin in HNC patients. Rêgo et al. (2017) [21] and Zhang et al. (2021) [23] showed that when metformin was used alone, cytotoxicity to HNSCC cells also decreased, thereby reducing cell viability by nearly 50% at higher doses. They also showed that metformin enhanced the proportion of apoptotic tumor cells most effectively at approximately 48 h post-treatment in selective cell lines (CAL27 = 25.4%, WSU-HN6 = 24.4%, SCC25 = 43.7%). The studies noted that metformin also significantly affects dose- and time-dependent responses by increasing G1/G0-phase cellular activity (P < 0.001). Furthermore, metformin effectively regulates protein expression levels by downregulating EGFR (epidermal growth factor receptor) activity in tumor cells (HSC3) via AMPK phosphorylation. In fact, after 24 to 48 h of administration of metformin in HNC patients, effects on associated proteins can be observed, and important cell cycle regulation, like cyclin-dependent kinases (CDKs), CDK inhibitors, mTOR, S6 Kinase, in HNSCC cells was found to be decreased among cyclins (D1, E) in cell lines.

Quality assessment

A total of five studies were evaluated across different domains of the AMSTAR 2.0 tool for assessing methodological robustness (Fig. 5). Overall, all included studies demonstrated a low level of confidence due to “no” responses in at least one of the critical domains- ‘protocol registration’.

Fig. 5.

Risk of Bias Assessment using AMSTAR 2.0 tool. Summary plot of the AMSTAR 2.0 quality assessment across all included systematic reviews, depicting judgments for each critical and non‑critical domain and showing overall low confidence due to frequent “no” responses in key domains such as protocol registration

Discussion

This umbrella review showed that metformin has promising potential to reduce the risk of HNC by altering the tumor microenvironment and sensitizing tumor cells to drugs and radiotherapy. Comprehensive analysis in this study also indicated that metformin enhances overall survival among users, particularly in DMT2, through dose-dependent mechanisms. Findings from the study characteristics indicated that the SRMAs mostly included primary evidence from the USA, China, and Taiwan, followed by Canada, Brazil, Korea, the UK, Saudi Arabia, and Italy. This suggests that there is an evidence gap between different geographical regions of the world. This can be attributed to structural reasons such as high burden of cancer, better research facilities, and easy availability of the epidemiological databases in these countries compared to others. According to bibliometric analysis Wang et al., in 2025 [25], the USA and China accounted for 54.24% of the global output on metformin and cancer-related research publications, mainly due to their strong research infrastructure, including funding. Surprisingly, with the increasing focus on oropharyngeal cancer, most of the studies included in this review focused on the larynx and oral cavities. In fact, studies by Clements et al., [26] in the same year reported that, with the increasing incidence of oropharyngeal cancer in HNC, the overall focus has shifted from larynx and oral cavity cancers. As the risk of HNC was not equal among males and females, a study by Wierzbicka et al., in 2025 [27] also noted that the risk of HNC cancer, particularly in the oral cavity (mainly salivary gland), risks was higher among women of all ages compared to laryngeal cancer where risk was more among men above 80 years of age group.

In addition, our findings suggested that metformin users have a lower risk of HNC incidence, and our meta-analysis also shows that metformin use can enhance survival probability in DMT2 patients. This can be attributed to the metformin’s ability to act like a barrier at three different levels- systemic, intracellular, and metabolic, thereby reducing its incidence. In a cohort study by Gaertner et al., in 2024 [28], similar observations were reported, suggesting that the 5-year survival probability was higher among metformin users than among non-users. This can be attributed to the drug’s ability to act intracellularly by disrupting cancer cells’ machinery, inhibiting mTOR (mechanistic target of rapamycin) signaling, and independently suppressing AMP-activated protein kinase activation. This observation was supported by Vander et al., [29], who found that, with the suppression of AMPK activation, metformin also leads to the activation of anti-tumorigenic cytokines by mediating cell cytotoxicity. It is noteworthy that in a study by Tsou et al., [30] it was found that patients with stage IV hypopharyngeal cancer showed poor effectiveness of the drug compared to early stages. The study noted that in later stages, intramolecular activities in cancerous cells were inhibited by metformin, thereby supporting its arrest in the G0/G1 phase, which helps limit the formation of new cancer cells. A similar observation was also noted by Rego et al., and Han et al., [20, 31]. suggesting that metformin activates Caspase 3, which promotes programmed cell death (apoptosis). In fact, the study also noted that metformin molecules limit CD24 + cell activity, thereby controlling tumor cell aggressiveness. At the metabolic level, these molecular activities promote the Warburg effect (the use of glycolysis for energy production rather than oxidative phosphorylation) by making cancer cells hypoxic and disrupting their survival. Studies by Kalender et al., Vazquez-Martin et al., Vander et al., and Stokes et al., [29, 32–34] also reported lower levels of available oxygen in aggressive cells. Metformin reduces Hypoxia-Inducible Factor-1 alpha (HIF-1α), a protective shield around tumor cells. This helps make cancer cells dysfunctional, thereby affecting their ability to spread and invade other tissues or organs, and to form new cancers at that site [35]. At the systemic level, a study conducted by Lee et al., in 2019 [36] suggested that metformin acts as an anti-neoplastic agent by reducing the increase in levels of insulin and insulin-like growth factor-1 (IGF-1), which signal cells to divide, thereby lowering the mitogenic activity of hyperinsulinemia. While metformin has demonstrated effectiveness in reducing incidence and improving survival among HNC patients at different levels, our analysis revealed differences in its effectiveness on DFS and LRR. Studies found improvements with lower sample sizes; however, pooled meta-analysis studies have shown insignificant results [12, 37, 38]. This can be attributed to the biological level improvements like reversing hypoxia-based radio-resistance among cancer cells, however, study level limitations like incomplete methodological like small sample sizes, and improper follow-up periods rather than a lack of therapeutic efficacy. Similar observation was also given by Jiao et al., (2022) [24] indicating more comprehensive evaluations were required to truly identify its effectiveness. In fact, in a study by Menendez et al., and Crist et al., [32, 39] that metformin did prevent pre-malignant conditions to form in the cellular region, however, high quality studies with larger sample size were required to estimate the power of the effectiveness. Additionally, meta-of meta-analysis also reflects on the evidence of publication bias in the included studies. This suggests potential existence of small study effect where studies used in performing this analysis may have disproportionately reflect trials with statistically significant outcomes while neglecting those with null findings indicating overemphasis of the overall positive result. However, this does not mean that evidence does not communicate meaningful information, rather it emphasizes on more careful interpretation of the overall findings of the study. Observations were also given by Dąbrowski Mariusz and Justo et al., in 2025 [40, 41] indicates that effect sizes influence pooled estimates, however, it requires more balanced and careful interpretation of the overall findings. Future research must prioritize the reporting of all clinical outcomes, regardless of significance, to ensure that meta-analytical syntheses are grounded in a reproducible and objectively accurate evidence base.

While metformin has its effectiveness at many different levels, evidences indicated that it can act as a low-cost, well-tolerated medication with minimal toxicity [14, 23]. However, a few studies indicate that metformin has failed to demonstrate beneficial effect on cancer survival. In a study by Lee et al., in 2019 [36], it was reported that metformin use did not support overall improvement in survival probability or disease-specific survival, local, regional, or distant, in HNC patients. This study notes a limitation by indicating that the patients included in their study were defined as metformin users based on use at the time of diagnosis, but the duration of use, dosage, medication compliance, and medication changes were not incorporated into the analysis, as this information was not available from the retrospective chart review. Furthermore, a similar observation was reported in the Suissa and Asoulay study [42], suggesting that without incorporating and adjusting for confounders, results tend to be affected by time-related biases. Nevertheless, there has been considerable variability in the included studies. Differences in the cancer subtypes, patient characteristics and dose administration of the metformin, may have produced heterogenous results. There were studies which mainly focused on oral cell squamous carcinoma of the oral cavity and larynx but few of the studies also addressed emerging subtypes like oropharyngeal carcinomas. This can lead to differences in the survival probability and incidence risk among the HNC diagnosed diabetic patients. Similar observation can be inferred from study by Tseng et al. (2014) and Huang et al. (2025) [43, 44] where the researchers reported that epidemiological outcomes vary depending upon the patient characteristics and types of HNC among diabetic patients due to differences in neoplastic activity and physiological responses. These studies also mention that patient’s physiological functioning also varies depending upon the age, sex, lifestyle factors, comorbidities and diabetic status in majority of the cases. Furthermore, it is important to also focus on the confounding implications of the covariates as presented in the studies. For instance, issue of indication bias among diabetic patients can lead to overestimation of drug related protective effects. This can be due to its preferential prescription to the patients with specific clinical profiles such as fewer comorbidities, early-stage diabetes and higher dosage of metformin. In a communication paper by Gaertner et al. (2024) [28], it was reported that metformin can have protective effect on HNC which can further be influenced by comorbidities and the initial tumor stage among patients, thereby presenting better outcomes in those consuming metformin compared to non-metformin users. Such inconsistencies and confounding factors across included studies presented in this review can also be one of the reasons for differences in the reported effect sizes and should be considered when interpreting the pooled findings. Moreover, the study notes that time bias can occur when the time prior to medication exposure is misclassified, resulting in skewing of the results toward risk reduction. In line with similar observations, Aldea et al. and He and Woldestron [45, 46] also note that such biases will limit the generalizability of the studies.

Strengths and limitations

This umbrella review exclusively assessed the highest level of available evidence to demonstrate the effectiveness of metformin in HNC. The review assessed its effectiveness in terms of overall survival probability, incidence risk, and cellular-level activity. This review fills the knowledge gap by adding evidence on how metformin improves overall health outcomes in individuals with HNC. However, the review also has certain limitations. Firstly, due to the heterogeneity of the available data, all variables indicating individual effectiveness were not analyzed. Secondly, the review focuses only on comparing DMT2 patients into two groups: those using metformin and those who were not; hence, the overall findings have limited generalizability to those who were not identified as DMT2-positive. Thirdly, number of included studies were only five with overall methodological robustness found to be low, this requires interpretation of the results with caution. Also, only three databases were used to retrieve relevant research records due to limitation in access to other databases. Furthermore, other side effects of metformin were not mentioned in any of the studies; this could be a potential effect modifier in this context. Lastly, this study focuses only on studies that demonstrate effectiveness in humans; hence, any evidence in the context of animals was not taken into consideration.

Conclusion

This review demonstrated the effectiveness of metformin to enhance overall survival in HNC patients. Metformin’s effectiveness indicates that DMT2 patients who consume metformin as an anti-glycemic drug have a reduced incidence of HNC. However, it is important to appreciate the limitation of the study result due to smaller number of studies forming evidence base and has weak methodological robustness. Therefore, future studies can focus on assessing how metformin affects non-DMT2 individuals and what the potential side effects of using the metformin drug in combination with radiotherapies are among cancer patients.

Authors’ contributions

V.M: Conceptualization, Methodology, Data curation, Formal analysis, Writing – original draft, Visualization . J.F.F : Methodology, Data curation, Validation, Writing – review & editing. A.M: Conceptualization, Supervision, Validation, Writing – review & editing, Project administration. M.B: Data curation, Investigation, Writing – review & editing. C.G: Validation, Writing – review & editing, Visualization . All authors have read and approved the final manuscript.

Funding

Open access funding provided by Dr. DY Patil Vidyapeeth, Pune (Deemed to be University). This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files. The protocol for this umbrella review was registered with PROSPERO (registration ID: CRD420261299771).

Declarations

Ethics approval and consent to participate

Not applicable. This umbrella review is based on previously published systematic reviews and meta-analyses and does not involve primary data collection from human subjects or animals.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.